The definition, in our words
The mean of the true positive rate and the true negative rate for a binary decision.
Why the distinction matters
The working draft on benchmarking uses balanced accuracy as the output behind a headline accuracy statement. It equals ordinary accuracy only when the test set is balanced.
How to use it in a comparison
Report the two rates beside the average, and state the class balance of the test set. An average can hide a system that is good with adults and poor with the children it is meant to protect.
Where it changes the evidence
A definition is not decoration. It decides which test is meaningful and which result can be compared. Where a supplier's material uses the term differently, the comparison should quote the supplier's use and place it beside this one. A difference in definition is a finding about comparability, and it belongs in the report rather than in a footnote.
The analysis units on this site use the term as the framework defines it. Where a unit turns the term into a question, the question is written on the unit page and the evidence artefact is named. The kit repeats both, so a row in the register and a paragraph in the report use the same meaning.
Source for this page
- ISO/IEC 27566-3 official record. The public record of the draft.
- ISO/IEC 27566-1:2025 official record. The published framework, free from ISO.
Retrieved 2026-10-21. This is our statement of the term, not a reproduction of the published definition.